This study will use newly collected, comprehensive data from rural Bangladesh to (a) investigate the impact of social networks on adult and elderly survival and (b) explore the variation in social networks amongst individuals. Its contributions include the following: First, unlike most previous studies which have used ad hoc, atheoretical, uni-dimensional summary network measures to represent social networks, this study will use a theoretically driven, empirically validated, multidimensional model of social networks. Specifically it will examine the relationship of network characteristics to survival within distinct social and kin roles (spouse, sons, daughters, brothers, sisters, other members of the residential compound (the bari), and confidant), recognizing that some roles may be more important than others. Second, it will address concerns about the universality of the relationship between social networks and adult survival across sociocultural settings by examining this issue in the context of a rural developing world setting. Third, the extensive sociodemographic information available will permit us to explore how different individual characteristics such as gender, age, income and education may modify the impact of social ties on survival, an area that has received inadequate attention thus far. Fourth, because of the wide variety of health status indicators available (self-reported and measured), we will be able to control for the confounding effect of prior health in a more comprehensive way than many studies. Fifth, this study will explore the extent to which social networks act as stressors versus buffers. Sixth, we will attempt to gain insight into the mechanisms through which social networks may affect survival, with a special focus on instrumental support exchanges. Finally, this study will use both a cross-sectional and a longitudinal approach to investigate how social networks vary across individuals, focusing on different conceptualizations and facets of networks.